Tensor Renormalization-Group study of the surface critical behavior of a frustrated two-layer Ising model

Basic Information

• Paper ID: 2510.21269
• Title: Tensor Renormalization-Group study of the surface critical behavior of a frustrated two-layer Ising model
• Author: C. Chatelain (Université de Lorraine, CNRS, LPCT, France)
• Classification: cond-mat.stat-mech (Statistical Mechanics)
• Publication Date: October 24, 2025 (arXiv preprint)
• Paper Link: https://arxiv.org/abs/2510.21269

Abstract

This paper investigates the surface critical behavior of two copies of two-dimensional Ising models coupled by frustrated spin-spin interactions. The interlayer coupling is known to be marginal, with bulk critical behavior belonging to the Ashkin-Teller (AT) universality class, identical to the $J_1$-$J_2$ Ising model. The author employs Tensor Renormalization-Group (TRG) numerical calculations to study surface critical behavior, extending the Bond-Weight Tensor Renormalization Group (BTRG) algorithm to systems with boundaries. The study reveals that the two-fold degeneracy of the surface magnetization scaling dimension in the AT model is lifted in the frustrated two-layer Ising model (F2LIM). This splitting is explained by the breaking of $\mathbb{Z}_2$ symmetry under spin reversal of individual Ising replicas in the F2LIM. The two distinct surface magnetization scaling dimensions $x_1^s$ and $x_2^s$ of the F2LIM satisfy a simple duality relation: $x_1^s = 1/4x_2^s$.

Research Background and Motivation

Problem to be Addressed

This paper studies the surface critical behavior of the frustrated $J_1$-$J_2$ Ising model and its equivalent model—the frustrated two-layer Ising model (F2LIM), representing a natural extension following investigations of bulk critical behavior.

Importance of the Problem

1. Theoretical Significance: Surface critical behavior is typically richer than bulk critical behavior. For the same bulk universality class, multiple surface universality classes may exist (ordinary, special, extraordinary, and surface phase transitions).
2. Methodological Significance: From the conformal field theory (CFT) perspective, surface critical behavior is described by unitary irreducible representations of a single Virasoro algebra, rather than the holomorphic and antiholomorphic algebras in the bulk.
3. Physical Controversy: Recent TRG calculations have questioned some classical Monte Carlo results for the $J_1$-$J_2$ Ising model, necessitating verification by more precise numerical methods.

Limitations of Existing Methods

1. Monte Carlo Methods: May suffer from systematic errors in frustrated systems
2. Small System Exact Diagonalization: Transfer matrix methods can only handle small-width strips
3. Existing TRG Algorithms: Primarily designed for periodic boundary conditions, lacking systematic methods for open and fixed boundary conditions

Research Motivation

The author previously constructed the F2LIM model in reference 8, which shares the same scaling limit as the $J_1$-$J_2$ Ising model but is more amenable to TRG calculations. This work extends the investigation to surface critical behavior, with particular focus on:

• Differences in surface critical behavior between F2LIM and the AT model
• Effects of symmetry breaking on surface scaling dimensions
• Obtaining more precise numerical results through larger system sizes

Core Contributions

1. Algorithm Extension: Extended the BTRG algorithm to systems with boundaries, capable of handling open boundary conditions (OBC) and fixed boundary conditions (FBC)
2. Precise Critical Line Determination: Obtained a more precise critical line for F2LIM (precision $O(10^{-5})$) through finite-size scaling analysis of domain wall free energy
3. Symmetry Breaking Discovery: Revealed that $\mathbb{Z}_2$ symmetry breaking at boundaries in F2LIM leads to the lifting of degeneracy in surface magnetization scaling dimensions
4. Duality Relation: Discovered that the two surface magnetization scaling dimensions satisfy the relation $x_1^s = 1/4x_2^s$, analogous to electromagnetic duality in Coulomb gases
5. Systematic Spectral Analysis: Provided complete low-energy spectral analysis of the AT model and F2LIM under different boundary conditions

Methodology Details

Task Definition

Input: Hamiltonian of a two-dimensional classical spin model and boundary conditions Output:

• Partition function and free energy
• Eigenvalue spectrum of the transfer matrix
• Scaling dimensions extracted via the gap-exponent relation

Constraints:

• System at or on the critical point/line
• Conformal invariance holds (second-order phase transition)

BTRG Boundary Algorithm Architecture

1. Tensor Decomposition Fundamentals

The system partition function is represented as a tensor network: $e^{-\beta H[s]} = T_{s_i s_j s_k s_l} T_{s_n s_m s_p s_q} \cdots$

where each rank-4 tensor $T$ corresponds to the Boltzmann weight at a lattice site, with four indices corresponding to spins at the four corners of the site.

2. Boundary Treatment

Rank-2 tensors $J_s$ are introduced at left and right boundaries, representing the Boltzmann weight of coupling between two surface spins.

3. Singular Value Decomposition (SVD)

Tensors are reconstructed as matrices and decomposed: $T_{s_1 s_2; s_3 s_4} = \sum_n U_{s_1 s_2; n} \sqrt{\Lambda_n} \times \sqrt{\Lambda_n} (V^+)_{n; s_3 s_4}$

The $\chi$ largest singular values are retained to control computational complexity.

4. BTRG Improvement

Using a modified SVD decomposition: $T_{s_1 s_2; s_3 s_4} = \sum_n U_{s_1 s_2; n} (\Lambda_n)^{1-k/2} (\Lambda_n)^k (\Lambda_n)^{1-k/2} (V^+)_{n; s_3 s_4}$

The optimal parameter is $k = -1/2$, significantly improving the precision of scaling dimension estimates.

5. Iterative Coarse-Graining

• First Step: Decompose tensors vertically and horizontally, forming rank-3 tensors
• Second Step: Contract 4-tensor and 8-tensor lattice sites, generating new rank-4 tensors
• Boundary Treatment: Left and right boundary tensors $T_L$ and $T_R$ are generated by contracting 3-tensor lattice sites

After $n$ iterations, each tensor corresponds to the Boltzmann weight of a $2^{n+1} \times 2^{n+1}$ spin block.

Scaling Dimension Extraction

Gap-Exponent Relation

The energy gap is computed from the transfer matrix eigenvalues $\Lambda_0 > \Lambda_1 > \cdots > \Lambda_n$: $E_n - E_0 = \ln(\Lambda_n/\Lambda_0)$

Periodic Boundary Conditions (PBC): $x_n = \frac{L}{2\pi}(E_n - E_0)$

Open/Fixed Boundary Conditions (OBC/FBC): $x_n^s = \frac{L}{\pi}(E_n - E_0)$

The factor difference arises from strip mapping to the upper half-plane (OBC/FBC) versus the entire complex plane (PBC).

Logarithmic Corrections for Marginal Operators

For marginal operators (e.g., 4-state Potts model): $x_n = \frac{L}{2\pi}(E_n - E_0) + \frac{d_n}{\ln L}$

Technical Innovations

1. Boundary Tensor Network Construction: Systematically handles left and right boundary tensors while maintaining consistency with bulk tensor iteration
2. Multiple Boundary Condition Types: Unified framework for OBC and various FBC types (identical/mixed)
3. Domain Wall Free Energy Method: Precisely locates critical points through the free energy difference $\Delta F$ between identical and mixed FBC, avoiding complex fitting for central charge extraction
4. High-Precision Computation: Retains $\chi=64$ states, with system sizes reaching $L=2^{10}$ spins

Experimental Setup

Model Definition

Ashkin-Teller Model

Hamiltonian: $-\beta H = J_2 \sum_{i,j} \sigma^A_{i,j}[\sigma^A_{i+1,j} + \sigma^A_{i,j+1}] + J_2 \sum_{i,j} \sigma^B_{i,j}[\sigma^B_{i+1,j} + \sigma^B_{i,j+1}]$$+ J_1 \sum_{i,j} \sigma^A_{i,j}\sigma^B_{i,j}[\sigma^A_{i+1,j}\sigma^B_{i+1,j} + \sigma^A_{i,j+1}\sigma^B_{i,j+1}]$

• Symmetry: Dihedral group $D_4$
• Critical line (self-dual line): $e^{-2J_1} = \sinh 2J_2$
• Critical exponent parametrization: $y \in [0, 3/2]$, via $\cos(\pi y/2) = \frac{1}{2}[e^{4J_1} - 1]$

Frustrated Two-Layer Ising Model (F2LIM)

Hamiltonian: $-\beta H = J_2 \sum_{i,j} \sigma^A_{i,j}[\sigma^A_{i+1,j} + \sigma^A_{i,j+1}] + J_2 \sum_{i,j} \sigma^B_{i,j}[\sigma^B_{i+1,j} + \sigma^B_{i,j+1}]$$+ J_1 \sum_{i,j} \sigma^A_{i,j}[\sigma^B_{i+1,j} - \sigma^B_{i,j+1}] + J_1 \sum_{i,j} \sigma^B_{i,j}[\sigma^A_{i+1,j} - \sigma^A_{i,j+1}]$

Key difference: Interlayer coupling is ferromagnetic in one direction and antiferromagnetic in the other (frustrated)

Evaluation Metrics

1. Scaling Dimension Precision: Deviation from exact values (AT model) or theoretical predictions
2. Critical Point Localization Precision: $O(10^{-5})$
3. Energy Level Identification: Correct classification of primary and descendant operators

Computational Parameters

• Retained States: $\chi = 16, 32, 64$
• Iteration Count: 6-10 iterations (corresponding to $L = 2^7$ to $2^{11}$)
• BTRG Parameter: $k = -1/2$
• Eigenvalue Solver: Arnoldi algorithm from ARPACK library

Boundary Condition Types

1. OBC: Open boundaries
2. Identical FBC: Same spin states at left and right boundaries
   • First group: $(\sigma^A, \sigma^B) = (\uparrow, \uparrow)$ on both sides
   • Second group: $(\sigma^A, \sigma^B) = (\uparrow, \downarrow)$ on both sides
3. Mixed FBC: Opposite spin states at left and right boundaries
   • First group: Left $(\uparrow, \uparrow)$, Right $(\downarrow, \downarrow)$
   • Second group: Left $(\uparrow, \downarrow)$, Right $(\downarrow, \uparrow)$

Experimental Results

AT Model Verification

Bulk Scaling Dimensions (PBC)

Scaling dimensions measured on the critical line (Figure 4 left):

• $x_\sigma = 1/8$ (two-fold degenerate): Magnetization density
• $x_{\sigma\tau} = 1/(8-4y)$: Polarization density
• $x_\varepsilon = 1/(2-y) = 4x_{\sigma\tau}$: Energy density
• $x_\sigma + 1$ (two-fold degenerate): First descendant of magnetization density

Precision Analysis:

• Near Ising point ($J_1 \approx 0$): Deviation from exact values <1%
• Near 4-state Potts point ($J_1 \approx 0.2$): Systematic deviations appear due to logarithmic corrections

Surface Scaling Dimensions (OBC)

Figure 4 right shows:

• $x_\sigma^s = 1/(2x_\varepsilon)$ (two-fold degenerate): Surface magnetization
• $x_{\sigma\tau}^s = 1$: Surface polarization
• $x_\sigma^s + 1$ (two-fold degenerate): First descendant

Good agreement with exact predictions from literature 20,21,22 ($J_1 \lesssim 0.15$).

Fixed Boundary Conditions

Identical FBC (Figure 5 left): Scaling dimensions $2, 4x_\sigma^s, 3, 4x_\sigma^s+1, 4$ observed

Mixed FBC (Figure 5 right): Only integer scaling dimensions $1, 2, 3, \ldots$ appear (descendants of identity operator)

F2LIM Critical Line Determination

Methodological Innovation

The critical point is determined through domain wall free energy $\Delta F$ (difference between identical and mixed FBC free energies):

• Paramagnetic Phase: $\Delta F$ decreases with iteration count (finite correlation length)
• Ferromagnetic Phase: $\Delta F$ diverges with iteration count (stable domain wall)
• Critical Point: Intersection of curves from different iteration counts (Figure 6)

Critical Line Results

Figure 7 shows the critical line in the $0 < J_1 < J_2$ region:

• Difference between $\chi=32$ and $\chi=64$ results: $O(10^{-4})$
• Difference from previous work 8 (central charge method): $O(10^{-4})$
• Tricritical point location: $J_1 \approx 0.425$ (estimated from $x_{\sigma\tau} = 1/8$)
• $J_1 > 0.425$: First-order phase transition region

F2LIM Scaling Dimension Analysis

Bulk Scaling Dimensions (PBC)

Figure 8 left shows:

• Lowest scaling dimension approaches $x_\sigma = 1/8$
• Key Finding: Second-lowest scaling dimension systematically deviates from the first, breaking two-fold degeneracy
• Third scaling dimension similar to AT's $x_{\sigma\tau}$
• Fitting $x_{\sigma\tau}$ yields parameter $y$, from which other scaling dimensions are derived

Surface Scaling Dimensions (OBC)—Core Finding

Figure 8 right presents the most important results:

Degeneracy Lifting: Two new energy levels appear without AT model counterparts

• Both originate from Ising point $x_\sigma^s = 1/2$
• Level splitting increases rapidly with $J_1$

Duality Relation: $x_2^s \approx \frac{1}{4x_1^s}$

Numerical verification is excellent (dashed line in Figure 8 right). This resembles electromagnetic duality in Coulomb gases $x_e(n) = n^2/(2g)$ and $x_m(m) = m^2 g/2$ (with $n=m=1$), but the coupling constant $g = 1/(2x_1^s) = x_2^s/2$ is smaller than the AT model's expected value $2/x_\varepsilon = x_\sigma^s$.

Other Energy Levels:

• Third scaling dimension agrees with bulk $x_\varepsilon = 4x_{\sigma\tau}$
• Still measurable in first-order phase transition region ($J_1 > 0.425$), but gap-exponent relation no longer applies

Fixed Boundary Conditions

First Identical FBC Group (Figure 9 left):

• Two lowest scaling dimensions: $2x_\sigma^s + 1$ and $4x_\sigma^s$ (degenerate at Ising point)
• Third: $2x_\sigma^s + 2$

First Mixed FBC Group (Figure 9 right):

• Two lowest scaling dimensions: $2x_\sigma^s$
• Third: $2x_\sigma^s + 1$
• Key Difference: No integer scaling dimensions appear (unlike AT model)

Second Identical FBC Group (Figure 10 left):

• Lowest: $2x_\varepsilon = 8x_{\sigma\tau}$
• Second (becomes third after $J_1 \gtrsim 0.25$): $4x_\sigma^s$
• Third (for $J_1 \lesssim 0.25$): Close to $3x_\varepsilon$, more likely $2x_\varepsilon + 1$ (first descendant)

Second Mixed FBC Group (Figure 10 right):

• Lowest: $x_\varepsilon$
• Third: $2x_\varepsilon$ or $x_\varepsilon + 1$
• Second: No clear explanation (Ising point is $2x_\sigma^s = 1$ but grows too rapidly)

Summary of Key Experimental Findings

1. Symmetry Breaking Mechanism: The AT model Hamiltonian is invariant under spin reversal of individual Ising replicas $(\sigma^A, \sigma^B) \to (-\sigma^A, \sigma^B)$ (four-spin coupling is even). F2LIM is only invariant under spin reversal combined with 90° rotation, but boundaries are inequivalent (two boundaries have fully ferromagnetic interlayer coupling, the other two fully antiferromagnetic), leading to boundary symmetry breaking.
2. Energy Spectrum Structure: The F2LIM OBC spectrum contains all major features of the AT model but adds two split surface magnetization modes.
3. Universality: Despite different microscopic symmetries, most of F2LIM's critical behavior (except surface magnetization degeneracy) remains consistent with the AT model.

Related Work

Frustrated Ising Model Research

1. Monte Carlo Methods:
   • Kalz et al. 4,6, Jin et al. 5,7: Support $J_1$-$J_2$ model belonging to AT universality class
   • Controversy: Phase transition order, tricritical point location
2. TRG Methods:
   • Li and Yang 1, Yoshiyama and Hukushima 2, Gangat 3: Question MC results
   • Chatelain 8: Construct F2LIM model, confirm AT universality class

Surface Critical Phenomena

1. Theoretical Framework:
   • Binder and Hohenberg 13, Cardy 14,16-18: General theory of surface critical phenomena
   • Pleimling 15: Review
2. AT Model Surface Behavior:
   • von Gehlen and Rittenberg 20,21: Small strip transfer matrix diagonalization
   • Alcaraz et al. 22: XXZ chain Bethe ansatz solution, exact surface exponents
3. TRG Boundary Algorithms:
   • Iino et al. 29,30: Boundary extensions of HoTRG and TNR
   • This work: Boundary extension of BTRG

Conformal Field Theory

1. Bulk CFT: Double Virasoro algebra, $c=1$ (AT and F2LIM)
2. Boundary CFT: Single Virasoro algebra, irreducible representations determine surface spectrum
3. Coulomb Gas: Electromagnetic duality relation $x_e x_m = 1/4$

Conclusions and Discussion

Main Conclusions

1. Algorithm Achievement: Successfully extended BTRG to systems with boundaries, capable of handling multiple boundary conditions, applicable to general 2D classical spin models
2. AT Model Verification:
   • Accurately reproduces known bulk and surface critical exponents
   • Verifies energy spectrum structure under different FBC
   • Confirms two-fold degeneracy of surface magnetization scaling dimension
3. F2LIM New Discoveries:
   • Precise critical line determination (precision $10^{-5}$)
   • Discovery of surface magnetization degeneracy lifting, producing two scaling dimensions $x_1^s$ and $x_2^s$
   • Establishment of duality relation $x_1^s = 1/(4x_2^s)$
   • Explanation as result of boundary $\mathbb{Z}_2$ symmetry breaking
4. Physical Mechanism: F2LIM's frustrated interlayer coupling (ferromagnetic in one direction, antiferromagnetic in the other) leads to inequivalent boundaries, breaking single-replica spin reversal symmetry

Limitations

1. Incomplete Theoretical Explanation:
   • Virasoro irreducible representations corresponding to $x_1^s, x_2^s$ not determined
   • Associated surface order parameters not identified
   • Coupling constant $g$ in duality relation differs from AT model expectation, lacking theoretical explanation
2. Numerical Precision Limitations:
   • Retained state number $\chi=64$ limits high excitation state precision
   • Logarithmic corrections near 4-state Potts point difficult to handle
   • Reliability of results in first-order phase transition region ($J_1 > 0.425$) questionable
3. Energy Spectrum Identification:
   • Second scaling dimension in second mixed FBC cannot be explained
   • Ambiguity whether some high excitation states are primary or descendant operators
4. System Size: Although reaching $L=2^{10}$, finite-size effects persist, particularly near critical line endpoints

Future Directions

1. Theoretical Analysis:
   • Determine irreducible representations of $x_1^s, x_2^s$ via CFT
   • Understand deep origins of duality relation
   • Construct corresponding surface order parameters
2. Numerical Improvements:
   • Increase retained state number $\chi$
   • Develop systematic methods for handling logarithmic corrections
   • Extend to other boundary conditions (free, periodic-mixed)
3. Model Generalization:
   • Study other frustrated multilayer models
   • Explore surface behavior of three-dimensional frustrated systems
   • Application to quantum frustrated models
4. Experimental Connection: Seek physical systems realizing F2LIM (e.g., magnetic thin films, cold atom systems)

In-Depth Evaluation

Strengths

1. Methodological Innovation:
   • BTRG boundary extension is systematic and practical, with clear algorithm flow in Figure 1
   • Domain wall free energy method for critical point location is elegant and effective, avoiding complex fitting
   • Unified framework handles multiple boundary conditions
2. Numerical Quality:
   • Comprehensive AT model verification with excellent agreement to exact values
   • Critical line precision reaches $10^{-5}$, superior to previous work
   • System sizes significantly larger than transfer matrix diagonalization methods
3. Physical Insights:
   • Symmetry breaking explanation clear and convincing
   • Duality relation $x_1^s = 1/(4x_2^s)$ is important discovery, suggesting deeper structure
   • Systematic comparison of AT and F2LIM reveals macroscopic effects of microscopic differences
4. Clear Presentation:
   • Rich figures (10 main plots) with high information content
   • Tables 1-2 summarize scaling dimensions for easy reference
   • Mathematical derivations and physical arguments well-balanced

Weaknesses

1. Theoretical Depth:
   • Duality relation lacks first-principles derivation
   • Coupling constant $g$ discrepancy insufficiently discussed
   • No exploration of connections to other integrable models (e.g., XXZ chain)
2. Experimental Design:
   • No systematic study of $\chi$ dependence (only three points: 16, 32, 64)
   • Lack of sensitivity analysis for different BTRG parameters $k$
   • Insufficient caution in first-order phase transition region data handling
3. Result Completeness:
   • Some energy levels unexplained (e.g., second mixed FBC)
   • Insufficient high excitation state analysis
   • No discussion of relationship between surface and bulk phase transitions
4. Reproducibility:
   • Lack of specific code or detailed implementation details
   • Some fitting procedures (e.g., parabolic fit for $x_{\sigma\tau}$) parameters not provided
   • Error estimation not systematic

Impact

1. Methodological Contribution:
   • BTRG boundary algorithm will become standard tool for studying 2D surface critical phenomena
   • Domain wall free energy method generalizable to other systems
   • Provides technical foundation for quantum systems (via quantum-classical mapping)
2. Physical Contribution:
   • Resolves long-standing questions about AT/F2LIM surface behavior
   • Duality relation may inspire new theoretical frameworks
   • Provides theoretical guidance for surface engineering of frustrated systems
3. Practical Value:
   • Algorithm directly applicable to magnetic material surface research
   • Provides reference for boundary effects in quantum computing
   • Critical line data serves as benchmark for other methods
4. Limitations:
   • Restricted to 2D classical systems
   • Incomplete theory limits deeper applications
   • Requires high-performance computing resources

Applicable Scenarios

1. Direct Application:
   • Surface critical behavior of 2D classical spin models
   • Phase transition research in frustrated magnetic systems
   • Critical phenomena in multilayer coupled systems
2. Potential Applications:
   • Quantum spin chains (via transfer matrix formalism)
   • Numerical verification of statistical field theory models
   • Numerical identification of boundary operators in CFT
3. Not Applicable:
   • Strongly coupled quantum systems (requires quantum TRG)
   • Three-dimensional systems (computational complexity too high)
   • Quantitative description of first-order transitions

Key References

8 C. Chatelain, PRE 111, 024109 (2025) - Author's previous work constructing F2LIM model

9 D. Adachi et al., PRB 105, L060402 (2022) - Original BTRG algorithm

22 F.C. Alcaraz et al., J. Phys. A 20, 6397 (1987) - Exact AT model surface exponents

29 S. Iino et al., PRB 100, 035449 (2019) - Boundary TRG methods (HoTRG version)

32 J. L. Cardy, J. Phys. A 17, L385 (1984) - Classical gap-exponent relation literature

Overall Assessment: This is a high-quality statistical mechanics numerical research paper with significant contributions in both methodology and physical discoveries. The BTRG boundary algorithm extension has broad application prospects, while the discovery of surface magnetization degeneracy lifting and the duality relation reveal new physics in frustrated systems. Although theoretical explanation remains incomplete, the numerical results are reliable and highly suggestive. The paper merits publication in top-tier physics journals and will have lasting impact on condensed matter theory and computational physics.